Artificial intelligence is the biggest talking point in technology right now. In fact, it’s probably the most important topic in the business world, too. It’s predicted to be a multi-trillion-pound industry in the coming years, perhaps being even more impactful than the advent of the World Wide Web in the 1990s. Yet, some believe that AI has been overhyped or, at best, misunderstood in what it can do. The legendary computer programmer Grady Booch is among those who claim AI is a glorified algorithm.
That said, it’s clear that AI is going to touch many areas of our economic, cultural, and social lives. Some of that disruption has already begun in areas like medicine, law, and the creative arts, even if the latter is proving controversial. However, we are also seeing an impact in sports and sports analysis, including sports predictions. This is an interesting use case for AI, not least because it arguably shows both its strengths and weaknesses.
Not a new science
The first thing to note is that AI, machine learning, and sports predictions are not a new combination. Supercomputers have been churning out predictions for years, decades even. As with Euro 2024 this summer, every major sports event tends to have “supercomputer predicts X outcome” articles picked up by tabloid newspapers. But the frequency of those articles down the years has built a consensus that they aren’t very good. Some of this is simply due to the nature of sports. We saw several supercomputers predict that Brazil would win the 2022 World Cup. Yet, Brazil flopped, and the eventual winner was Argentina, a team rarely spat out by the AI number-crunching.
None of this should be surprising, of course. If we look at something like horse racing betting, it makes sense that the odds are set through the judgment of the fastest and strongest horses. Data analysis can make things more interesting. What happens if X horse races against Y for the first time? How will this horse handle the rain? – Yet any racing fan will tell you that a horse can do inexplicable things, including refusing to race. You simply cannot model an animal’s brain.
If the previous statements are understood, i.e., that animals and humans can do unpredictable things that make sports difficult to accurately predict, then what is the point of using AI in the first place? The answer, broadly, lies in the fact that its usefulness is not supposed to be found in viewing it as a perfect model. An AI sports prediction model is supposed to tell us the most likely outcome based on available data. That’s different from telling us what will happen.
Human predictions remain valid
Keeping that last part in mind shows the usefulness of AI sports predictions compared to human ones. We can read a handful of data points – a football team’s form, previous results between two teams, and so on. AI can look at and compare millions of them. We humans and the AI are doing the same thing, but the latter is doing it a much vaster scale. And that’s the rub: the nature of sports remains the same, so there is never going to be a magic answer regardless of the amount of data analyzed.
While AI offers significant accuracy improvements in predictions, it lacks the nuanced understanding and intuition that human experts bring to the table – this is because humans understand humans better, and sport is a human endeavor. Thus, the best approach combines AI’s analytical power with human intuition and experience. Ultimately, AI is a valuable tool, but not a replacement for the human element in sports predictions.
